CANCER IMAGE ANALYSIS
Figure 2: Photometric stereo reconstruction of a mucosal surface. Rows 1-2: six images are taken with a single camera and multiple light sources. The resulting surface reconstruction is shown row 3 (left: height map, right: 3D surface rendering).
using stereovision, structure from motion and structured light have been proposed, photometric stereo [17, 18], which uses a single camera and multiple light sources to reconstruct 3D surface topography has promise due to its real-time reconstruction and high resolution. An example is shown in Figure 2. By taking advantage of the shape information, it is possible to combine some advantages of the 3D shape-based analysis of CT colonography CAD with the colour and texture-based analysis of EndoCAD approaches. However, performing precise 3D reconstruction of a highly reflective mucosal surface in a narrow confined space poses difficult engineering challenges.
88
In CT colonography, there is considerable interest in developing CAD for reduced, or even non-cathartically prepared patients [19], as this preparation is seen as a factor contributing to lower patient engagement rates. CAD systems designed for this application must differentiate lesions from residual waste in the colon, which is a difficult problem even with the use faecal tagging agents [20]. Extension of existing CAD techniques to alternative imaging such as low-dose CT [21] or dual-energy CT [22] have been proposed. Recently, there has been notable research activity on further resolving polyps from non-polyps by taking advantage of both CT series (prone and supine) through spatial registration
[23, 24]. Non-polyps such as stool are mobile, and move considerably relative to the colon when the patient is repositioned in the scanner. Finally, there is growing research activity in going beyond pure detection, and into computer-aided diagnosis and treatment. The classification of a polyp as hyperplastic or adenomatous can affect the surveillance interval for patient follow-up. Typically, this diagnosis is done by a histopathologist who examines excised tissue under a microscope. However, recent studies [25] have argued for an in-vivo classification during colonoscopy to reduce healthcare costs. The diagnosis can be achieved using a variety of imaging techniques, for
Volume 9 Issue 3 • July/August 2014